Happy path / edge case

The happy path is the run where the user does the expected thing and every system cooperates: valid input, fast network, results found, payment approved. Everything else splits into two piles that get confused constantly. Alternate and error paths are ordinary and expected: the card is declined, the search finds nothing, the user cancels, the invite is already pending. Edge cases are the unusual ones out at the boundaries and extremes: zero rows, ten thousand rows, a 200-character company name, a token that expires mid-request. The demo you record lives on the happy path; the support tickets live everywhere else.

The productive move is to enumerate the edges by category instead of waiting to discover them. Zero (empty list, no results, first run), one, and many (a thousand rows, pagination, a name that wraps four lines). Extremes (a 200-character company name, a 40MB upload, an emoji-only search). Time (expired token, timezone rollover, a request that lands after the user navigated away). Permissions, offline, duplicate submit, and the double-click that fires the mutation twice.

Two traps. First, edge cases are not rare in aggregate: individually a 1 percent case, together they are most of your one-star reviews. Second, AI tools ship the happy path by default, so a generated feature usually renders beautifully with three seeded rows and falls apart on zero rows or a failed fetch. Name the edges in the prompt or you will find them in production.